Musings of a Chief Analytics Officer: Do Companies Need an AI Strategy?
by Soumendra Mohanty & Sachin Vyas
“AI has transformational capabilities." “Everybody wants to do AI.”
These have become moot points. The important question now is, “What is relevant for your enterprise?”
There is a growing trend on the part of the popular press, platform vendors, and researchers to suggest that in terms of trends, everything is AI. During the course of writing the book How to Compete in the Age of Artificial Intelligence, we spoke with many C-level executives, practitioners, and researchers, and we realized that beyond the theoretical definition of AI - from an “AI for the enterprise” perspective - there is no common answer for, “What is AI and what is not?”
In our view, there are 3 classes of AI:
1. AI as UI (Ubiquitous Intelligence): There is very little left in our day-to-day life that is not impacted by AI in some way. Everything we see around us is increasingly powered by an AI algorithm, and while AI is gaining a status of ubiquity, we are sure that we have only scratched the surface regarding what AI can mean for us.
2. AI as AAAI (Assisted, Augmented, Autonomous Intelligence): Generally speaking, an intelligent system is something that processes information in order to do something purposeful. Whether you are searching for something on the Internet, shopping online, or seeking a medical diagnosis, intelligent systems are likely to play a key role in the process. These intelligent systems seek the best plan of action to accomplish their assigned goals (as assisting capabilities, augmented capabilities, and autonomous capabilities.) You can think of these three roles of intelligent systems as a maturity continuum.
- Assisted intelligence: Primarily focused on improving what people and organizations are already doing in their day-to-day activities.
- Augmented intelligence: Helps people and organizations with its complementary capabilities, so that they can do things they couldn’t otherwise do on their own.
- Autonomous intelligence: Creates and deploys machines that are intelligent and adaptive by nature so that they can act on their own, completely eliminating human involvement.
3. AI as II (Invisible Interface): We’ve started talking to our machines—not with commands, menus, and frantic keystrokes, but by using human language. Natural language processing has seen incredible advances and now we don’t need to be a machine in order to communicate with another machine.
To separate the hype from reality, and to be candid, we've heard references to AI as if it is a single entity. In reality, it is not. Perhaps the most simplistic way to refer to AI is that it is a collection of technologies (Deep Learning, Natural Language Processing, Image Recognition, Voice Recognition, Reinforcement Learning, Question Answering Machines, Adversarial Training, and Automation.) These are all separate disciplines, though they can work well together in examples like Watson, Echo/Alexa, Google Home, Tesla Autonomous Cars, intelligent navigation systems like Waze, Amazon recommendation systems, Nest Thermostats, and more. Beyond these sleek examples, integration of these different technologies is not easy. The biggest challenge is figuring out how to make AI real for enterprises to solve every day problems.
Businesses get carried away with the above examples and start forming “AI first” strategies for their enterprise with a notion that they can implement something similar for their LOBs. Getting inspired is one thing, but doing a lift-and-shift and hoping that the same magic will happen can be a foolish attempt. During our conversations with many business executives and strategists, we found a common pattern about their AI strategy: take X (current products, services, customer base, geographic plans, employee skills) and simply add AI to it.
Now, how do we identify AI use cases? Is there a formula? When we researched and double-clicked on every single AI use case, starting from the most famous to obscure yet relevant ones, we found that there is something common: almost all AI use cases address three fundamental tenets: functionality, training data and algorithms, and an end-user need.
Using these three tenets, a use case can be described based on the formula:
Deliver (functionality) by learning from (data), thus fulfilling (user need.)
Not all use cases are equal from a business outcome perspective, so there can’t be a one-size-fits-all strategy.
When identifying AI opportunities for your enterprise it’s important to differentiate and seriously question whether you are going to get incremental improvement vs. whether you are going to create fundamentally new capabilities. Both are equally important, and each will lead to a different strategy: horizontal vs. vertical.
The horizontal strategy is to make an AI offering that is generic enough to solve a host of common problems and can do the same things cheaper, faster, and better. This includes generalized image, video, speech, data processing, automation, platforms, and more, using the familiar MLaaS model.
In contrast, the vertical strategy is all about data dominance and domain depth. Here, AI is not an incremental add to what the enterprise already does. Rather, AI becomes the core to unlocking totally new opportunities. It is not easy: you will need data - lots of data. You will also need SME/domain expertise, willing customers with whom you can co-create, skills (data science,) and above all, you will need tolerance for failure. It is a tall order for sure.
In the book How to Compete in the Age of Artificial Intelligence, you will find many answers to the questions that the above might have raised, in addition to many other topics like AIOps, IoT, automation, cybersecurity, and developing a human-machine integrated strategy.
About the Author
Soumendra Mohanty is an acclaimed thought leader and SME in the areas of analytics, IoT, AI, cognition, and automation. He is a two- decade veteran with expertise in next-gen Big Data solutions, BI architectures, the enterprise data warehouse, customer insight solutions, and industry-specific advanced analytics solutions. With his broad experience, he has designed and implemented data analytics solutions for Fortune 500 clients across industry verticals. Soumendra is an advisor with the Harvard Business Review Advisory Council. He is also associated with the Indian Statistical Institute and various universities as a visiting faculty member specializing in Big Data and analytics. Soumendra speaks at various global forums, CAO advisory forums, and educational institutions. He is author of several books including Big Data Imperatives (Apress).
Sachin Vyas is an entrepreneur with over 20 years of experience in technology, data, and analytics. He was the founder and CEO of AugmentIQ Data Sciences, a company acquired by LTI. LTI is a global technology consulting and digital solutions company. AugmentIQ focused on creating platforms and solutions specific to Big Data engineering and data sciences and provided solutions to large and complex data and analytics problems for financial services companies. He currently heads LTI’s platforms and enabling solutions with converging technologies across devices, data, computing, and artificial intelligence. Sachin is the recipient of the EMC Transformer Awards-2012 and the SKOCH Digital Inclusion Awards-2012 in India. He has a mechanical engineering degree from VNIT, Nagpur, India.
This article was contributed by Soumendra Mohanty & Sachin Vyas, authors of How to Compete in the Age of Artificial Intelligence.